Learning Temporal Consistency Correlation Filter with Gradient Feature for UAV Tracking

This paper presents a novel target tracking model that integrates a learning temporal consistency constraint (LTCF) guided by gradient features, seamlessly embedding it within a correlation filter-based tracking framework. This model incorporates the gradient feature of the correlation filter to enhance information extraction and feature representation in the tracking model for small targets. Additionally, the entire model is effectively solved using an alternating optimization method. To assess the performance of the proposed tracking algorithm, we conduct comparative experiments on LTCF and other tracking models using multiple datasets. The experimental results demonstrate that the LTCF model delivers robust tracking performance in complex scenes, ensuring both accuracy and reliability in target tracking.